Multi-objective optimization framework for generative design of horseshoe-shaped pipe arrangement in pre-stressed underground bundles

Abstract

The underground bundle composite pipe integrated by transverse pre-stressing (UBIT) is an emerging method for underground excavation. It offers advantages such as the enhanced structural performance, reduced deformation, simplified construction process, and improved economic efficiency. However, the complexity inherent in the UBIT structure, coupled with the subjective nature of individual experience, renders the manual design of the UBIT pipe arrangement within tunnel cross-sections a challenging task. To address this, this research proposes an intelligent design method named multi-objective optimization-based generative design (MOOGD), which combines multi-objective optimization principles with generative design techniques. The proposed MOOGD has two main innovations: (a) It enables an “end-to-end” design process that enhances efficiency while minimizing human errors; (b) It incorporates five classic multi-objective optimization algorithms, offering flexibility for engineers to select or substitute algorithms based on specific project needs. The effectiveness of MOOGD is validated through a case study of the Pingli Station construction project on Shanghai Metro Line 20. It is a two-layer underground station in Shanghai Metro Line 20, with a horseshoe-shaped circular tunnel cross-section. Compared to the original engineering design, the MOOGD-based UBIT pipe arrangement design solutions using five candidate multi-objective optimization algorithms achieve significant improvements of 11.03 %, 10.64 %, 10.95 %, 10.89 %, and 10.80 %, respectively. MOOGD provides a highly efficient and reliable design tool for engineers, facilitating a rapid exploration of design alternatives and the automated generation of optimal UBIT pipe arrangements that adhere to real-world engineering specifications.